mutisya/w2v-bert-adapters-14lang-e10-25_52-v8

W2V-BERT 2.0 ASR Adapters

This repository contains per-language bottleneck adapters for automatic speech recognition (ASR) trained on top of facebook/w2v-bert-2.0.

Model Description

• Base Model: facebook/w2v-bert-2.0 (600M parameters, frozen)
• Adapter Architecture: Bottleneck adapters (Pfeiffer-style, dim=64)
• Decoder: Lightweight transformer decoder (2 layers)
• Training: CTC loss with extended vocabulary for double vowels

Trained Adapters

Training in progress...

Adapter	Language	WER	Train Samples

Successful: 14/14

✅ Good (WER < 30%): 9 swh_Latn_v2: 4.49% swh_Latn_salt: 13.60% kik_Latn: 16.41% luo_Latn: 16.58% swh_Latn_v1: 17.50% eng_Latn_tts: 22.24% eng_Latn_salt: 24.80% lug_Latn_salt: 27.48% ach_Latn: 29.28%

⚡ Medium (30-60% WER): 2 kam_Latn: 31.20% mer_Latn: 36.34%

⚠️ Poor (60-90% WER): 1 nyn_Latn: 65.37%

❌ Collapsed (WER >= 90%): 2 ful_Latn: 100.00% teo_Latn: 99.87%

Architecture

The model uses:

1. Frozen w2v-bert-2.0 encoder - Extracts audio representations
2. Bottleneck adapters - Language-specific adaptation (trainable)
3. Lightweight decoder - Transformer decoder blocks (trainable)
4. LM head - Per-language vocabulary projection (trainable)

Usage

Each adapter folder contains:

• adapter_weights.pt - Bottleneck adapter weights
• decoder_weights.pt - Decoder block weights
• lm_head_weights.pt - Language model head weights
• final_norm_weights.pt - Final layer norm weights
• vocab.json - Language-specific vocabulary
• adapter_config.json - Adapter configuration
• metrics.json - Training metrics

Training Configuration

• Epochs: 10
• Base Learning Rate: 0.0005 (adaptive based on dataset size)
• Batch Size: 48 x 1
• Extended Vocabulary: True
• Adapter Dimension: 64

License

Apache 2.0